Molecular Intelligence Simplifies Cryo-EM Data Analysis with AI

A team of scientists from Purdue University’s College of Science and the Rosen Center for Advanced Computing has established Molecular Intelligence, a software company dedicated to helping researchers determine the 3D structures of biomolecules using cryogenic-electron microscopy (cryo-EM).

Leading the company is Daisuke Kihara, a Professor of Biological Sciences and Computer Science, who is also affiliated with the Purdue Institute for Cancer Research and the Purdue Institute for Drug Discovery. The founding team includes Charles Christoffer, a Senior Computational Scientist at the Rosen Center for Advanced Computing, and Genki Terashi, an Assistant Research Scientist in the Department of Biological Sciences.

Molecular Intelligence serves researchers in bioengineering, medical science, and the pharmaceutical industry who use cryo-EM to analyze the 3D structures of proteins, nucleic acids, and other biomolecules. These structures provide critical insights into protein function and play a key role in drug development by guiding the design of molecules that interact with and modify protein activity.

Founded in the summer of 2024, Molecular Intelligence secured an exclusive license from the Purdue Innovates Office of Technology Commercialization to sell its software in January 2025.

The Rise and Challenges of Cryo-EM

According to Kihara, academic institutions, pharmaceutical firms, and biotech companies worldwide are increasingly investing in cryo-EM technology, with more than 1,100 locations now equipped with cryo-EM systems. This technology has become the primary method for determining most protein structures.

However, Kihara highlights a major challenge: cryo-EM produces low-resolution image data, making it difficult to accurately model protein, nucleic acid, and drug molecule structures.

“Current data processing, including structure model building, can take many days, sometimes stretching into weeks, which adds significant costs. When automated methods fail to support structure modeling, researchers are forced to manually build models using interactive software—a process that is both time-consuming and prone to errors. If a drug target, such as a protein, is modeled incorrectly, it can severely impact downstream research, including drug design.”

Daisuke Kihara, Professor, Purdue University

Molecular Intelligence’s Solution

To address these challenges, Kihara, Christoffer, and Terashi have developed a deep learning-powered software suite designed to streamline cryo-EM data analysis.

“Our software uses deep learning to identify atom positions in low-resolution cryo-EM maps. From there, it constructs accurate 3D models of proteins and DNA/RNA. It’s fully automated, runs on a standard workstation with a moderate GPU, and requires no special training to use. For researchers who prefer hands-on support, we also offer modeling services.”

Daisuke Kihara, Professor, Purdue University

With this technology, Molecular Intelligence aims to make cryo-EM data processing faster, more accessible, and more reliable for researchers tackling complex biomolecular structures.